Alibaba Unveils Its AI Ambitions with New Models

At its annual Apsara conference, Alibaba introduced a suite of new Qwen AI models alongside what it called China’s most powerful AI processor and an agentic smartphone solution. The company stated that its Qwen 4 model is currently undergoing training and revealed that its forthcoming Qwen 4.5 and Qwen 5 series are projected to scale to 5 trillion and 10 trillion parameters, respectively. In comparison, Alibaba’s current flagship, the Qwen3.8-Max, features 2.4 trillion parameters.

In conjunction with the Qwen strategic plan, Alibaba introduced new multimodal models encompassing speech, audio, and vision capabilities. It also launched Qwen Intelligence, an agent platform designed for smartphones capable of managing complex, cross-application tasks. As competition intensifies within the AI hardware sector, Alibaba noted that its upcoming Zhenwu V900 AI processor can deliver triple the performance of its predecessor. It is slated for commercial release in the first quarter of 2027, with Alibaba asserting the chip will be the most powerful in China.

The company further emphasized advancements in recursive self-enhancement. In a semiconductor design experiment, the Qwen3.8-Max model spent over 60 hours optimizing its output, executing more than 10,000 electronic design automation tool calls. Alibaba reported that this process reduced chip area by 42% without degrading performance. The firm’s pursuit of larger models comes as it targets increasingly sophisticated AI workloads and progresses toward what it characterizes as the Machine Intelligence era.

Group CEO Eddie Wu projected that machines would eventually generate over 1,000 times the cognitive output of all humanity, an increase from less than 3% today. To achieve this, Wu explained that the company is targeting over 20GW of global Alibaba Cloud data center capacity by 2032 as it prepares to meet the industry’s exponentially growing demand for AI. Joe Tsai, Chairman of Alibaba Group, added that the company intends to transition AI from technological breakthroughs toward genuine value creation.